AI Is Using Your Dealership's Pages Without Saying Your Name. Two Studies Put It at 40% to 62% of Citations.

Open a browser, ask Google's AI Mode which dealer near you has a specific trim in stock, and read the answer. Then look at the source links stacked underneath it. There is a real chance your dealership's website is sitting in that source list while your dealership's name appears nowhere in the sentence the shopper actually reads.

That gap has a name now, and as of this week it has numbers behind it. Search Engine Land published a piece on July 29 called "Ghost citations: Why AI search cites your content, not your brand." The headline finding: across roughly 16 million brand appearances in a recent 30-day window of AI answers, "around 40% of AI citations didn't name the source brand in the answer."

The engine-by-engine split is the part dealers should read twice, because two of the worst offenders are the two surfaces where American car shoppers start:

• Perplexity: 52%
• Google AI Mode: 49%
• Google AI Overviews: 41%
• ChatGPT: 37%
• Gemini: 25%
• Grok: 22%
• Microsoft Copilot: 19%

The definitions are simple. A citation gets recorded when the answer links to your page. A mention gets recorded when your name appears in the text the shopper reads. A ghost citation is the first one without the second one. Your content did the work. Somebody else, or nobody, got the credit.

The number is contested. The phenomenon is not.

Be careful with that 40%, and be careful with anyone who quotes it at you as gospel. The data came from Writesonic, and the column's co-author Samanyou Garg is disclosed in the piece as Writesonic's founder and CEO. It is a company that sells AI visibility software publishing a study about how badly you need AI visibility software. The column does not link a published methodology.

So we went looking for whether anyone independent had measured the same thing. Somebody had, seven weeks earlier, and the answer came back worse.

Semrush, working with Kevin Indig of Growth Memo, published a ghost citations study on June 9 built on 3,981 domain appearances across 115 prompts, run against ChatGPT, Google AI Overviews, Google AI Mode and Gemini in 14 countries. Their headline: "62% of citations don't lead to brand mentions." The breakdown was 61.7% cited but never named, 13.2% both cited and named, and 25.1% named without being cited at all.

Two vendors, two samples, two different definitions, two different engine mixes, and a range from 40% to 62%. That spread is worth stating plainly rather than papering over. Nobody has a clean number yet. The direction is not in dispute by anyone measuring it: somewhere between four and six of every ten times an AI answer leans on a business's page, the business stays anonymous.

The two datasets also disagree about which engine is worst. Both agree Gemini names brands most often (Writesonic put its ghost rate lowest at 25%, Semrush measured an 83.7% mention rate). But Writesonic ranks Perplexity worst at 52%, while Semrush's ChatGPT figures, a 20.7% mention rate against an 87% citation rate, would put ChatGPT at the bottom. Treat the leaderboard as unsettled and the pattern as real.

One more limitation before we get to your store: the Semrush sample spans 14 countries without breaking out a US-only figure, and neither study published results for local or automotive queries. What follows is our read on how a measured cross-industry pattern lands on a franchised dealership, which is analysis rather than a study of dealer sites.

Dealers have already lived through this once

Strip the AI language out and franchised dealers will recognize the shape of this immediately.

For twenty years, dealers have handed their inventory, pricing, and photography to third-party listing sites, and those sites built enormous businesses ranking for the searches shoppers actually type. Your VIN, your reconditioning, your photos, and the shopper's memory of the whole experience attached to the aggregator's brand instead of yours. Dealers spent a decade fighting that, building their own SRP and VDP experiences specifically to own the moment of discovery.

A ghost citation is the same trade running at machine speed, without a contract, without a monthly invoice, and without a rep you can call.

There is a second piece of history worth knowing, because it explains how the industry got caught out. Indig first named the ghost citation problem back on April 20. Meanwhile every AI visibility product built in the last eighteen months shipped with a citation counter on the dashboard, because citations are trivial to count. The industry standardized on the easy metric before checking whether it was the right one. Call it the backlink-count era all over again, when agencies sold link volume for years while Google was quietly grading something else.

Citations and mentions barely track each other

Here is the finding that should reset how you evaluate any AI visibility report crossing your desk.

Indig and Semrush ran a much larger study, published July 20, covering 1,094 US categories in ChatGPT with five buyer-intent prompts each, monthly from January through June 2026, across more than 220,000 domains, 50,000 brands and 600,000 citations. Two results from it:

Only 21% of the most-cited domains in a category were also the most-mentioned brand. The correlation between citations and mentions came in at -0.229, which is to say slightly negative. Search Engine Land's independent write-up reported that same 21% finding and concluded that "traditional SEO metrics aren't enough to explain who owns a topic."

Sit with that for a second. The domain getting cited most in a category is usually a different domain than the brand getting named most. If your vendor's monthly report leads with a citation count, there is a real possibility it is trending the number that runs slightly opposite to the number that decides whether a shopper hears your name.

We have written before about why a single AI visibility score can't be trusted and about how fast a citation you earned can disappear. This is a different failure. It says a citation can be perfectly real, perfectly stable, and still worth close to nothing to you.

What this means for your dealership

Your service and how-to content is the most exposed thing you publish. Semrush broke results out by query type, and the split is stark. Informational prompts produced an 89.3% citation rate and only an 18% mention rate. Comparative prompts produced a 43.3% mention rate, 2.4 times the informational rate. How-to landed at 42.8%, commercial at 35.6% against an 84.4% citation rate. Now think about what fills a typical dealer blog: "how often should you rotate your tires," "what does a lease residual mean," "signs your brakes need service." That library is precisely the format that gets harvested anonymously.

The content that earns your name is the content almost no dealer publishes. Comparison and buying-decision material carries the highest mention rates in the data. Very few dealership websites have a real "2026 Silverado versus F-150 for towing" page, a lease-versus-finance breakdown for their own state, or a trim-by-trim comparison written for a shopper rather than for a crawler. The gap between what dealers publish and what gets a brand named is wide, and it is fixable this quarter.

Your name is missing from your own sentences. Go pull up your busiest service page right now and read the body copy. Odds are it says "we," "our dealership," "our certified technicians," and "contact us today." The store's legal name lives in the header, the footer, and the schema, and never once appears next to an actual claim. Writesonic's third recommendation is to pair brand names directly with claims rather than distantly, and Semrush's guidance is to add clearer brand references. A model reading "our technicians are factory-certified" has nothing to attach that claim to. Give it "[Your Store Name]'s technicians are factory-certified" and it does.

The engines that ghost the most are the ones your shoppers use. Perplexity's 52% makes for a good headline, and Perplexity sends a rounding error of traffic to a car dealer. Google AI Mode at 49% and AI Overviews at 41% sit inside the search results where the overwhelming majority of American vehicle shopping still begins. That is where the anonymity actually costs you a lead.

Nobody owns your category yet, and that will not last. In the 1,094-category study, only 15.2% of categories had a clear owner, 31.2% had an emerging leader, and 53.7% were completely unsettled, with no brand appearing in even three of five prompts. Indig calculated that 89.3% of estimated AI search demand sits in categories with no clear owner, and that the biggest categories by volume were the least owned, 11.3% versus 19.0% for smaller ones. The flip side is the part that matters: once a brand did clearly own a category, it held first place in 90.4% of month-over-month comparisons. Ownership is rare and hard to get, and durable once you have it. That study measured brand categories rather than local dealer queries, so read it as direction rather than as a forecast for "Ford dealer near me." The direction says the land is still open.

What to do about it

1. Make your vendor report mentions, separately from citations, per engine. Two numbers, one row per engine. If the product cannot separate them, you now know what it is actually measuring.
2. Read ten real answers, not the dashboard. Ask Google AI Mode and ChatGPT the ten questions your shoppers actually ask, including model comparisons, trade-in, financing, and "best dealer near me." Note every answer where your site is a source and your name is absent. That list is your work queue.
3. Put the dealership name inside the claim. Sweep your service, financing, trade-in, and about pages so the store name sits in the same sentence as the specific claim, not just in the template furniture around it. Do this without turning the copy into keyword mush, because a human still has to read it.
4. Build comparison content on purpose. Model versus model, trim versus trim, lease versus finance, new versus certified. The data says comparative and how-to formats are more than twice as likely to get a brand named as generic informational posts.
5. Publish something only you could publish. Writesonic's second recommendation is proprietary named research, and for a dealer that is more achievable than it sounds. Your local market pricing, your service turnaround times, your inventory mix by trim. A number that exists nowhere else has to be attributed to somebody.
6. Stop paying for citation volume. If a proposal promises a number of citations, it is quoting the metric with the -0.229 correlation. We wrote about the vendors selling manufactured citations earlier this summer, and this data is one more reason that pitch belongs in the trash.

The fair version of the argument

For a news publisher, a ghost citation is close to theft, and pulling the content is a defensible response. For a dealership it is a missed introduction. The shopper still got useful information about a vehicle you sell, in your market, sourced from your page. Nobody told them where it came from.

That is fixable by writing differently rather than by hiding. The engines are already reading your pages. The work is making sure the sentence they lift has your name welded to it.

If you want a second set of eyes on which of your pages are getting cited without being named, and what your service and comparison content would need to change, book a time with us. Finding out takes an afternoon. Watching a competitor become the answer to your market's questions takes about two quarters, and it does not announce itself.